NVIDIA Confidential Computing Expands with New Bare Metal Option for Secure AI Workloads

Secure Scaling Across Global Infrastructure

NVIDIA has expanded its Confidential Computing portfolio with the launch of Confidential Bare Metal, a new deployment option designed to support frontier model builders operating on sensitive workloads. This solution enables organizations to run proprietary models and datasets on dedicated third-party infrastructure while maintaining full control over their intellectual property and data security.

Converged Security Architecture

The updated framework now offers three distinct deployment paths: Confidential Bare Metal for dedicated hardware environments, Confidential Containers for Kubernetes-based workflows, and Confidential Virtual Machines for virtualized setups. These options allow enterprises to align their security posture with specific operational and compliance requirements across hybrid and multi-cloud ecosystems.

Additionally, NVIDIA is integrating Confidential Computing into its next-generation Grace Blackwell and Rubin platforms, along with the NVIDIA DSX OS. This integration establishes a reference architecture for cloud providers aiming to build secure AI factories capable of handling high-value training and inference tasks globally.

Key enhancements include advanced platform security controls that safeguard data in transit, verify system integrity, and enforce hardware-level access policies. These features ensure that even on shared infrastructure, sensitive AI workloads remain isolated and protected from unauthorized access.

Industry adoption is already underway. Cohere plans to embed Confidential Bare Metal into its Model Vault offering, leveraging trusted execution environments to extend hardware-rooted encryption directly to GPU memory and interconnects. Similarly, Google Cloud is bringing support to its A4X Max instances powered by NVIDIA GB300 NVL72 systems, combining raw compute performance with cryptographic safeguards for model IP and training data.

Why it matters for GPU / AI infrastructure: As demand grows for training and deploying large-scale AI models, securing data and models across distributed GPU clusters becomes critical. Confidential Bare Metal provides a scalable path for enterprises to leverage external compute resources without compromising on privacy or regulatory compliance—especially relevant for organizations operating in regulated sectors such as finance, healthcare, and government.